質問編集履歴
4
コードの統一
test
CHANGED
|
File without changes
|
test
CHANGED
|
@@ -28,8 +28,6 @@
|
|
|
28
28
|
|
|
29
29
|
img = None
|
|
30
30
|
|
|
31
|
-
label = []
|
|
32
|
-
|
|
33
31
|
for i in range(len(old_train)):
|
|
34
32
|
|
|
35
33
|
if np.isin(old_train[i][1], mask_list):
|
|
@@ -60,8 +58,6 @@
|
|
|
60
58
|
|
|
61
59
|
img = None
|
|
62
60
|
|
|
63
|
-
label = []
|
|
64
|
-
|
|
65
61
|
for i in range(len(old_test)):
|
|
66
62
|
|
|
67
63
|
if np.isin(old_test[i][1], mask_list):
|
|
@@ -70,266 +66,258 @@
|
|
|
70
66
|
|
|
71
67
|
img = old_train[i][0]
|
|
72
68
|
|
|
73
|
-
|
|
69
|
+
label = old_train[i][1]
|
|
74
70
|
|
|
75
71
|
count += 1
|
|
76
72
|
|
|
77
73
|
else:
|
|
78
74
|
|
|
75
|
+
img = np.vstack((img, old_test[i][0]))
|
|
76
|
+
|
|
77
|
+
label = np.hstack((label, old_test[i][1]))
|
|
78
|
+
|
|
79
|
+
|
|
80
|
+
|
|
81
|
+
test = tuple_dataset.TupleDataset(img, label)
|
|
82
|
+
|
|
83
|
+
```
|
|
84
|
+
|
|
85
|
+
学習部分も含めたコードは以下になります.
|
|
86
|
+
|
|
87
|
+
```python
|
|
88
|
+
|
|
89
|
+
def main():
|
|
90
|
+
|
|
91
|
+
gane = [64, 64, 128, 128, 256, 256, 256, 512, 512, 512, 512, 512, 512]
|
|
92
|
+
|
|
93
|
+
|
|
94
|
+
|
|
95
|
+
# 抽出するデータラベル
|
|
96
|
+
|
|
97
|
+
mask_list = [4, 30, 55, 72, 95, 1, 32, 67, 73, 91]
|
|
98
|
+
|
|
99
|
+
|
|
100
|
+
|
|
101
|
+
batchsize = 32
|
|
102
|
+
|
|
103
|
+
epoch = 300
|
|
104
|
+
|
|
105
|
+
learnrate = 0.05
|
|
106
|
+
|
|
107
|
+
gpu = 0
|
|
108
|
+
|
|
109
|
+
out = "result"
|
|
110
|
+
|
|
111
|
+
|
|
112
|
+
|
|
113
|
+
print('Using CIFAR100 dataset.')
|
|
114
|
+
|
|
115
|
+
# クラス数を設定
|
|
116
|
+
|
|
117
|
+
class_labels = len(mask_list)
|
|
118
|
+
|
|
119
|
+
# cifar100を全データ格納
|
|
120
|
+
|
|
121
|
+
old_train, old_test = get_cifar100()
|
|
122
|
+
|
|
123
|
+
|
|
124
|
+
|
|
125
|
+
# 学習用データを抽出
|
|
126
|
+
|
|
127
|
+
count = 0
|
|
128
|
+
|
|
129
|
+
img = None
|
|
130
|
+
|
|
131
|
+
for i in range(len(old_train)):
|
|
132
|
+
|
|
133
|
+
if np.isin(old_train[i][1], mask_list):
|
|
134
|
+
|
|
135
|
+
if count == 0:
|
|
136
|
+
|
|
137
|
+
img = old_train[i][0]
|
|
138
|
+
|
|
139
|
+
label = old_train[i][1]
|
|
140
|
+
|
|
141
|
+
count += 1
|
|
142
|
+
|
|
143
|
+
else:
|
|
144
|
+
|
|
79
145
|
img = np.vstack((img, old_train[i][0]))
|
|
80
146
|
|
|
81
|
-
|
|
147
|
+
label = np.hstack((label, old_train[i][1]))
|
|
148
|
+
|
|
149
|
+
|
|
150
|
+
|
|
82
|
-
|
|
151
|
+
train = tuple_dataset.TupleDataset(img, label)
|
|
152
|
+
|
|
153
|
+
|
|
154
|
+
|
|
155
|
+
# 評価用データを抽出
|
|
156
|
+
|
|
157
|
+
count = 0
|
|
158
|
+
|
|
159
|
+
img = None
|
|
160
|
+
|
|
161
|
+
for i in range(len(old_test)):
|
|
162
|
+
|
|
163
|
+
if np.isin(old_test[i][1], mask_list):
|
|
164
|
+
|
|
165
|
+
if count == 0:
|
|
166
|
+
|
|
167
|
+
img = old_train[i][0]
|
|
168
|
+
|
|
83
|
-
label
|
|
169
|
+
label = old_train[i][1]
|
|
170
|
+
|
|
171
|
+
count += 1
|
|
172
|
+
|
|
173
|
+
else:
|
|
174
|
+
|
|
175
|
+
img = np.vstack((img, old_test[i][0]))
|
|
176
|
+
|
|
177
|
+
label = np.hstack((label, old_test[i][1]))
|
|
84
178
|
|
|
85
179
|
|
|
86
180
|
|
|
87
181
|
test = tuple_dataset.TupleDataset(img, label)
|
|
88
182
|
|
|
183
|
+
|
|
184
|
+
|
|
185
|
+
model = L.Classifier(VGG(gane, class_labels))
|
|
186
|
+
|
|
187
|
+
|
|
188
|
+
|
|
189
|
+
if gpu >= 0:
|
|
190
|
+
|
|
191
|
+
# Make a specified GPU current
|
|
192
|
+
|
|
193
|
+
chainer.cuda.get_device_from_id(gpu).use()
|
|
194
|
+
|
|
195
|
+
model.to_gpu() # Copy the model to the GPU
|
|
196
|
+
|
|
197
|
+
|
|
198
|
+
|
|
199
|
+
optimizer = chainer.optimizers.MomentumSGD(learnrate)
|
|
200
|
+
|
|
201
|
+
optimizer.setup(model)
|
|
202
|
+
|
|
203
|
+
optimizer.add_hook(chainer.optimizer.WeightDecay(5e-4))
|
|
204
|
+
|
|
205
|
+
|
|
206
|
+
|
|
207
|
+
train_iter = chainer.iterators.SerialIterator(train, batchsize)
|
|
208
|
+
|
|
209
|
+
test_iter = chainer.iterators.SerialIterator(test, batchsize,
|
|
210
|
+
|
|
211
|
+
repeat=False, shuffle=False)
|
|
212
|
+
|
|
213
|
+
# Set up a trainer
|
|
214
|
+
|
|
215
|
+
updater = training.StandardUpdater(train_iter, optimizer, device=gpu)
|
|
216
|
+
|
|
217
|
+
trainer = training.Trainer(updater, (epoch, 'epoch'), out=out)
|
|
218
|
+
|
|
219
|
+
|
|
220
|
+
|
|
221
|
+
# Evaluate the model with the test dataset for each epoch
|
|
222
|
+
|
|
223
|
+
trainer.extend(extensions.Evaluator(test_iter, model, device=gpu))
|
|
224
|
+
|
|
225
|
+
|
|
226
|
+
|
|
227
|
+
# Reduce the learning rate by half every 25 epochs.
|
|
228
|
+
|
|
229
|
+
trainer.extend(extensions.ExponentialShift('lr', 0.5),
|
|
230
|
+
|
|
231
|
+
trigger=(25, 'epoch'))
|
|
232
|
+
|
|
233
|
+
|
|
234
|
+
|
|
235
|
+
# Dump a computational graph from 'loss' variable at the first iteration
|
|
236
|
+
|
|
237
|
+
# The "main" refers to the target link of the "main" optimizer.
|
|
238
|
+
|
|
239
|
+
trainer.extend(extensions.dump_graph('main/loss'))
|
|
240
|
+
|
|
241
|
+
|
|
242
|
+
|
|
243
|
+
# Take a snapshot at each epoch
|
|
244
|
+
|
|
245
|
+
trainer.extend(extensions.snapshot(), trigger=(epoch, 'epoch'))
|
|
246
|
+
|
|
247
|
+
|
|
248
|
+
|
|
249
|
+
# Write a log of evaluation statistics for each epoch
|
|
250
|
+
|
|
251
|
+
trainer.extend(extensions.LogReport())
|
|
252
|
+
|
|
253
|
+
|
|
254
|
+
|
|
255
|
+
trainer.extend(extensions.PrintReport(
|
|
256
|
+
|
|
257
|
+
['epoch', 'main/loss', 'validation/main/loss',
|
|
258
|
+
|
|
259
|
+
'main/accuracy', 'validation/main/accuracy', 'elapsed_time']))
|
|
260
|
+
|
|
261
|
+
|
|
262
|
+
|
|
263
|
+
trainer.run()
|
|
264
|
+
|
|
265
|
+
|
|
266
|
+
|
|
267
|
+
if __name__ == '__main__':
|
|
268
|
+
|
|
269
|
+
main()
|
|
270
|
+
|
|
89
271
|
```
|
|
90
272
|
|
|
273
|
+
|
|
274
|
+
|
|
91
|
-
|
|
275
|
+
エラー文は以下になります.
|
|
92
|
-
|
|
93
|
-
```python
|
|
94
|
-
|
|
95
|
-
def main():
|
|
96
|
-
|
|
97
|
-
gane = [64, 64, 128, 128, 256, 256, 256, 512, 512, 512, 512, 512, 512]
|
|
98
|
-
|
|
99
|
-
|
|
100
|
-
|
|
101
|
-
# 抽出するデータラベル
|
|
102
|
-
|
|
103
|
-
mask_list = [4, 30, 55, 72, 95, 1, 32, 67, 73, 91]
|
|
104
|
-
|
|
105
|
-
|
|
106
|
-
|
|
107
|
-
batchsize = 32
|
|
108
|
-
|
|
109
|
-
epoch = 300
|
|
110
|
-
|
|
111
|
-
learnrate = 0.05
|
|
112
|
-
|
|
113
|
-
gpu = 0
|
|
114
|
-
|
|
115
|
-
out = "result"
|
|
116
|
-
|
|
117
|
-
|
|
118
|
-
|
|
119
|
-
print('Using CIFAR100 dataset.')
|
|
120
|
-
|
|
121
|
-
# クラス数を設定
|
|
122
|
-
|
|
123
|
-
class_labels = len(mask_list)
|
|
124
|
-
|
|
125
|
-
# cifar100を全データ格納
|
|
126
|
-
|
|
127
|
-
old_train, old_test = get_cifar100()
|
|
128
|
-
|
|
129
|
-
|
|
130
|
-
|
|
131
|
-
# 学習用データを抽出
|
|
132
|
-
|
|
133
|
-
count = 0
|
|
134
|
-
|
|
135
|
-
img = None
|
|
136
|
-
|
|
137
|
-
label = []
|
|
138
|
-
|
|
139
|
-
for i in range(len(old_train)):
|
|
140
|
-
|
|
141
|
-
if np.isin(old_train[i][1], mask_list):
|
|
142
|
-
|
|
143
|
-
if count == 0:
|
|
144
|
-
|
|
145
|
-
img = old_train[i][0]
|
|
146
|
-
|
|
147
|
-
label = old_train[i][1]
|
|
148
|
-
|
|
149
|
-
count += 1
|
|
150
|
-
|
|
151
|
-
else:
|
|
152
|
-
|
|
153
|
-
img = np.vstack((img, old_train[i][0]))
|
|
154
|
-
|
|
155
|
-
label = np.hstack((label, old_train[i][1]))
|
|
156
|
-
|
|
157
|
-
|
|
158
|
-
|
|
159
|
-
train = tuple_dataset.TupleDataset(img, label)
|
|
160
|
-
|
|
161
|
-
|
|
162
|
-
|
|
163
|
-
# 評価用データを抽出
|
|
164
|
-
|
|
165
|
-
count = 0
|
|
166
|
-
|
|
167
|
-
img = None
|
|
168
|
-
|
|
169
|
-
label = []
|
|
170
|
-
|
|
171
|
-
for i in range(len(old_test)):
|
|
172
|
-
|
|
173
|
-
if np.isin(old_test[i][1], mask_list):
|
|
174
|
-
|
|
175
|
-
if count == 0:
|
|
176
|
-
|
|
177
|
-
img = old_train[i][0]
|
|
178
|
-
|
|
179
|
-
#label = old_train[i][1]
|
|
180
|
-
|
|
181
|
-
count += 1
|
|
182
|
-
|
|
183
|
-
else:
|
|
184
|
-
|
|
185
|
-
img = np.vstack((img, old_train[i][0]))
|
|
186
|
-
|
|
187
|
-
#label = np.hstack((label, old_train[i][1]))
|
|
188
|
-
|
|
189
|
-
label += old_train[i][1]
|
|
190
|
-
|
|
191
|
-
|
|
192
|
-
|
|
193
|
-
test = tuple_dataset.TupleDataset(img, label)
|
|
194
|
-
|
|
195
|
-
|
|
196
|
-
|
|
197
|
-
model = L.Classifier(VGG(gane, class_labels))
|
|
198
|
-
|
|
199
|
-
|
|
200
|
-
|
|
201
|
-
if gpu >= 0:
|
|
202
|
-
|
|
203
|
-
# Make a specified GPU current
|
|
204
|
-
|
|
205
|
-
chainer.cuda.get_device_from_id(gpu).use()
|
|
206
|
-
|
|
207
|
-
model.to_gpu() # Copy the model to the GPU
|
|
208
|
-
|
|
209
|
-
|
|
210
|
-
|
|
211
|
-
optimizer = chainer.optimizers.MomentumSGD(learnrate)
|
|
212
|
-
|
|
213
|
-
optimizer.setup(model)
|
|
214
|
-
|
|
215
|
-
optimizer.add_hook(chainer.optimizer.WeightDecay(5e-4))
|
|
216
|
-
|
|
217
|
-
|
|
218
|
-
|
|
219
|
-
train_iter = chainer.iterators.SerialIterator(train, batchsize)
|
|
220
|
-
|
|
221
|
-
test_iter = chainer.iterators.SerialIterator(test, batchsize,
|
|
222
|
-
|
|
223
|
-
repeat=False, shuffle=False)
|
|
224
|
-
|
|
225
|
-
# Set up a trainer
|
|
226
|
-
|
|
227
|
-
updater = training.StandardUpdater(train_iter, optimizer, device=gpu)
|
|
228
|
-
|
|
229
|
-
trainer = training.Trainer(updater, (epoch, 'epoch'), out=out)
|
|
230
|
-
|
|
231
|
-
|
|
232
|
-
|
|
233
|
-
# Evaluate the model with the test dataset for each epoch
|
|
234
|
-
|
|
235
|
-
trainer.extend(extensions.Evaluator(test_iter, model, device=gpu))
|
|
236
|
-
|
|
237
|
-
|
|
238
|
-
|
|
239
|
-
# Reduce the learning rate by half every 25 epochs.
|
|
240
|
-
|
|
241
|
-
trainer.extend(extensions.ExponentialShift('lr', 0.5),
|
|
242
|
-
|
|
243
|
-
trigger=(25, 'epoch'))
|
|
244
|
-
|
|
245
|
-
|
|
246
|
-
|
|
247
|
-
# Dump a computational graph from 'loss' variable at the first iteration
|
|
248
|
-
|
|
249
|
-
# The "main" refers to the target link of the "main" optimizer.
|
|
250
|
-
|
|
251
|
-
trainer.extend(extensions.dump_graph('main/loss'))
|
|
252
|
-
|
|
253
|
-
|
|
254
|
-
|
|
255
|
-
# Take a snapshot at each epoch
|
|
256
|
-
|
|
257
|
-
trainer.extend(extensions.snapshot(), trigger=(epoch, 'epoch'))
|
|
258
|
-
|
|
259
|
-
|
|
260
|
-
|
|
261
|
-
# Write a log of evaluation statistics for each epoch
|
|
262
|
-
|
|
263
|
-
trainer.extend(extensions.LogReport())
|
|
264
|
-
|
|
265
|
-
|
|
266
|
-
|
|
267
|
-
trainer.extend(extensions.PrintReport(
|
|
268
|
-
|
|
269
|
-
['epoch', 'main/loss', 'validation/main/loss',
|
|
270
|
-
|
|
271
|
-
'main/accuracy', 'validation/main/accuracy', 'elapsed_time']))
|
|
272
|
-
|
|
273
|
-
|
|
274
|
-
|
|
275
|
-
trainer.run()
|
|
276
|
-
|
|
277
|
-
|
|
278
|
-
|
|
279
|
-
if __name__ == '__main__':
|
|
280
|
-
|
|
281
|
-
main()
|
|
282
276
|
|
|
283
277
|
```
|
|
284
278
|
|
|
285
|
-
|
|
279
|
+
---------------------------------------------------------------------------
|
|
280
|
+
|
|
286
|
-
|
|
281
|
+
ValueError Traceback (most recent call last)
|
|
282
|
+
|
|
283
|
+
<ipython-input-4-eb3fae69fa84> in <module>()
|
|
284
|
+
|
|
285
|
+
224
|
|
286
|
+
|
|
287
|
+
225 if __name__ == '__main__':
|
|
288
|
+
|
|
289
|
+
--> 226 main()
|
|
290
|
+
|
|
291
|
+
|
|
292
|
+
|
|
293
|
+
<ipython-input-4-eb3fae69fa84> in main()
|
|
294
|
+
|
|
295
|
+
145 print(len(label))
|
|
296
|
+
|
|
297
|
+
146 print(label.ndim)
|
|
298
|
+
|
|
299
|
+
--> 147 train = tuple_dataset.TupleDataset(img, label)
|
|
300
|
+
|
|
301
|
+
148
|
|
302
|
+
|
|
287
|
-
|
|
303
|
+
149 count = 0
|
|
304
|
+
|
|
305
|
+
|
|
306
|
+
|
|
307
|
+
/usr/local/lib/python3.6/dist-packages/chainer/datasets/tuple_dataset.py in __init__(self, *datasets)
|
|
308
|
+
|
|
309
|
+
35 if len(dataset) != length:
|
|
310
|
+
|
|
311
|
+
36 raise ValueError(
|
|
312
|
+
|
|
313
|
+
---> 37 'dataset of the index {} has a wrong length'.format(i))
|
|
314
|
+
|
|
315
|
+
38 self._datasets = datasets
|
|
316
|
+
|
|
317
|
+
39 self._length = length
|
|
318
|
+
|
|
319
|
+
|
|
320
|
+
|
|
321
|
+
ValueError: dataset of the index 1 has a wrong length
|
|
288
322
|
|
|
289
323
|
```
|
|
290
|
-
|
|
291
|
-
---------------------------------------------------------------------------
|
|
292
|
-
|
|
293
|
-
ValueError Traceback (most recent call last)
|
|
294
|
-
|
|
295
|
-
<ipython-input-4-eb3fae69fa84> in <module>()
|
|
296
|
-
|
|
297
|
-
224
|
|
298
|
-
|
|
299
|
-
225 if __name__ == '__main__':
|
|
300
|
-
|
|
301
|
-
--> 226 main()
|
|
302
|
-
|
|
303
|
-
|
|
304
|
-
|
|
305
|
-
<ipython-input-4-eb3fae69fa84> in main()
|
|
306
|
-
|
|
307
|
-
145 print(len(label))
|
|
308
|
-
|
|
309
|
-
146 print(label.ndim)
|
|
310
|
-
|
|
311
|
-
--> 147 train = tuple_dataset.TupleDataset(img, label)
|
|
312
|
-
|
|
313
|
-
148
|
|
314
|
-
|
|
315
|
-
149 count = 0
|
|
316
|
-
|
|
317
|
-
|
|
318
|
-
|
|
319
|
-
/usr/local/lib/python3.6/dist-packages/chainer/datasets/tuple_dataset.py in __init__(self, *datasets)
|
|
320
|
-
|
|
321
|
-
35 if len(dataset) != length:
|
|
322
|
-
|
|
323
|
-
36 raise ValueError(
|
|
324
|
-
|
|
325
|
-
---> 37 'dataset of the index {} has a wrong length'.format(i))
|
|
326
|
-
|
|
327
|
-
38 self._datasets = datasets
|
|
328
|
-
|
|
329
|
-
39 self._length = length
|
|
330
|
-
|
|
331
|
-
|
|
332
|
-
|
|
333
|
-
ValueError: dataset of the index 1 has a wrong length
|
|
334
|
-
|
|
335
|
-
```
|
3
エラー文の追加
test
CHANGED
|
File without changes
|
test
CHANGED
|
@@ -292,17 +292,29 @@
|
|
|
292
292
|
|
|
293
293
|
ValueError Traceback (most recent call last)
|
|
294
294
|
|
|
295
|
-
<ipython-input-
|
|
295
|
+
<ipython-input-4-eb3fae69fa84> in <module>()
|
|
296
|
-
|
|
296
|
+
|
|
297
|
-
22
|
|
297
|
+
224
|
|
298
|
-
|
|
298
|
+
|
|
299
|
-
22
|
|
299
|
+
225 if __name__ == '__main__':
|
|
300
|
-
|
|
300
|
+
|
|
301
|
-
--> 22
|
|
301
|
+
--> 226 main()
|
|
302
|
+
|
|
303
|
+
|
|
304
|
+
|
|
302
|
-
|
|
305
|
+
<ipython-input-4-eb3fae69fa84> in main()
|
|
303
|
-
|
|
304
|
-
|
|
306
|
+
|
|
305
|
-
1
|
|
307
|
+
145 print(len(label))
|
|
308
|
+
|
|
309
|
+
146 print(label.ndim)
|
|
310
|
+
|
|
311
|
+
--> 147 train = tuple_dataset.TupleDataset(img, label)
|
|
312
|
+
|
|
313
|
+
148
|
|
314
|
+
|
|
315
|
+
149 count = 0
|
|
316
|
+
|
|
317
|
+
|
|
306
318
|
|
|
307
319
|
/usr/local/lib/python3.6/dist-packages/chainer/datasets/tuple_dataset.py in __init__(self, *datasets)
|
|
308
320
|
|
2
エラー文全体を表示
test
CHANGED
|
File without changes
|
test
CHANGED
|
@@ -288,6 +288,22 @@
|
|
|
288
288
|
|
|
289
289
|
```
|
|
290
290
|
|
|
291
|
+
---------------------------------------------------------------------------
|
|
292
|
+
|
|
293
|
+
ValueError Traceback (most recent call last)
|
|
294
|
+
|
|
295
|
+
<ipython-input-1-b519d4e5f5fa> in <module>()
|
|
296
|
+
|
|
297
|
+
223
|
|
298
|
+
|
|
299
|
+
224 if __name__ == '__main__':
|
|
300
|
+
|
|
301
|
+
--> 225 main()
|
|
302
|
+
|
|
303
|
+
|
|
304
|
+
|
|
305
|
+
1 frames
|
|
306
|
+
|
|
291
307
|
/usr/local/lib/python3.6/dist-packages/chainer/datasets/tuple_dataset.py in __init__(self, *datasets)
|
|
292
308
|
|
|
293
309
|
35 if len(dataset) != length:
|
1
情報追加
test
CHANGED
|
File without changes
|
test
CHANGED
|
@@ -6,8 +6,92 @@
|
|
|
6
6
|
|
|
7
7
|
|
|
8
8
|
|
|
9
|
+
データ抽出部分のみのコードは以下になります.
|
|
10
|
+
|
|
9
11
|
```python
|
|
10
12
|
|
|
13
|
+
# 抽出するデータラベル
|
|
14
|
+
|
|
15
|
+
mask_list = [4, 30, 55, 72, 95, 1, 32, 67, 73, 91]
|
|
16
|
+
|
|
17
|
+
|
|
18
|
+
|
|
19
|
+
# cifar100を全データ格納
|
|
20
|
+
|
|
21
|
+
old_train, old_test = get_cifar100()
|
|
22
|
+
|
|
23
|
+
|
|
24
|
+
|
|
25
|
+
# 学習用データを抽出
|
|
26
|
+
|
|
27
|
+
count = 0
|
|
28
|
+
|
|
29
|
+
img = None
|
|
30
|
+
|
|
31
|
+
label = []
|
|
32
|
+
|
|
33
|
+
for i in range(len(old_train)):
|
|
34
|
+
|
|
35
|
+
if np.isin(old_train[i][1], mask_list):
|
|
36
|
+
|
|
37
|
+
if count == 0:
|
|
38
|
+
|
|
39
|
+
img = old_train[i][0]
|
|
40
|
+
|
|
41
|
+
label = old_train[i][1]
|
|
42
|
+
|
|
43
|
+
count += 1
|
|
44
|
+
|
|
45
|
+
else:
|
|
46
|
+
|
|
47
|
+
img = np.vstack((img, old_train[i][0]))
|
|
48
|
+
|
|
49
|
+
label = np.hstack((label, old_train[i][1]))
|
|
50
|
+
|
|
51
|
+
|
|
52
|
+
|
|
53
|
+
train = tuple_dataset.TupleDataset(img, label)
|
|
54
|
+
|
|
55
|
+
|
|
56
|
+
|
|
57
|
+
# 評価用データを抽出
|
|
58
|
+
|
|
59
|
+
count = 0
|
|
60
|
+
|
|
61
|
+
img = None
|
|
62
|
+
|
|
63
|
+
label = []
|
|
64
|
+
|
|
65
|
+
for i in range(len(old_test)):
|
|
66
|
+
|
|
67
|
+
if np.isin(old_test[i][1], mask_list):
|
|
68
|
+
|
|
69
|
+
if count == 0:
|
|
70
|
+
|
|
71
|
+
img = old_train[i][0]
|
|
72
|
+
|
|
73
|
+
#label = old_train[i][1]
|
|
74
|
+
|
|
75
|
+
count += 1
|
|
76
|
+
|
|
77
|
+
else:
|
|
78
|
+
|
|
79
|
+
img = np.vstack((img, old_train[i][0]))
|
|
80
|
+
|
|
81
|
+
#label = np.hstack((label, old_train[i][1]))
|
|
82
|
+
|
|
83
|
+
label += old_train[i][1]
|
|
84
|
+
|
|
85
|
+
|
|
86
|
+
|
|
87
|
+
test = tuple_dataset.TupleDataset(img, label)
|
|
88
|
+
|
|
89
|
+
```
|
|
90
|
+
|
|
91
|
+
学習部分も含めたコードは以下になります.
|
|
92
|
+
|
|
93
|
+
```python
|
|
94
|
+
|
|
11
95
|
def main():
|
|
12
96
|
|
|
13
97
|
gane = [64, 64, 128, 128, 256, 256, 256, 512, 512, 512, 512, 512, 512]
|